I like building things that actually work in production — secure, tested, well-documented.

• 🎓 B.Tech Computer Engineering (2025 Graduate)
• 📍 From India
• 🚧 Currently: building full-stack projects, learning system design ,learning AI .....basically learning things and implementing them.
• 💡 Interests: Web Development, API Design, Problem Solving, Open Source

Live demo →

A full-stack observability tool: apps ship logs to it via API, it tracks error rates in real time, fires threshold-based alerts, and uses an LLM to turn raw log spikes into a plain-English root-cause summary.

• Backend: Node.js + Express 5, PostgreSQL (Supabase), raw SQL with window functions for analytics
• AI integration: Groq + Llama 3.3 70B for natural-language log analysis — provider-agnostic service layer, easy to swap models
• Security: JWT auth, bcrypt, per-resource ownership checks on every endpoint (tested: user A cannot read user B's data), rate limiting, Helmet.js headers, Zod validation
• Frontend: React + Vite + Recharts for real-time dashboards
• Quality: 26 passing tests (Jest + Supertest), Dockerized, deployed on Render

Node.js React PostgreSQL Groq/Llama 3 JWT Docker

Live demo → · API docs (Swagger) →

A REST API built like it had to go to production, not just pass a demo — because that's the standard I hold my own code to.

• Security: JWT auth with bcrypt hashing, route-level authorization, tiered rate limiting (global + stricter on auth), Helmet.js headers, Joi validation on every input
• API design: Full CRUD with pagination, filtering, and sorting; documented with Swagger/OpenAPI so it's usable without reading the source
• Observability: Structured logging with Winston + Morgan (request audit trail, error stack traces)
• Quality: 19 passing tests (Jest + Supertest), one-command Docker Compose setup, centralized error handling

Node.js Express PostgreSQL JWT Swagger Docker

An agent that answers natural-language questions about any GitHub account or repo — built to understand how AI agents actually work under the hood, not just call an API and print the response.

• Agent loop: the LLM (via Groq) decides which tool to call, Python executes the real GitHub API request, results feed back to the LLM, which responds in natural language — with full conversation memory for follow-ups
• Tools implemented: repo stats (stars, issues, language), account-level aggregation (total stars, top repo, most-used language), and repo listing sorted by popularity
• Stack: Python, Groq API (gpt-oss-20b) for function-calling, Flask backend with session memory, GitHub REST API

Python Groq Flask LLM Function-Calling

I keep my fundamentals sharp alongside shipping projects — 500+ problems solved on LeetCode, 5★ on HackerRank, 1450 rating on CodeChef.